[Mlir-commits] [mlir] [MLIR][XeGPU] Add distribution pattern for xegpu load & store matrix from sg to wi (PR #183179)

[Nishant Patel]

================
@@ -489,6 +490,137 @@ struct SgToWiMultiDimReduction
   }
 };
 
+/// Helper to compute distributed coordinates for matrix ops.
+/// When not using subgroup_block_io, each workitem computes its own
+/// coordinates based on the layout and lane ID.
+static SmallVector<Value> computeDistributedCoordsForMatrixOp(
----------------
nbpatel wrote:

@Jianhui-Li are you ok with this?

https://github.com/llvm/llvm-project/pull/183179